A New Aaect-perceiving Interface and Its Application to Personalized Music Selection 1 Background and Motivation

A wearable computer that perceives and responds to the wearer's aaective state ooers a new kind of perceptual interface. Instead of asking the user to continuously select preferences from a menu, the aaective wearable gets to know its wearer's preferences by recognizing and responding to signals that carry emotional information. This paper presents a new design for an aaective wearable interface, aimed at applications that do not require extensive access to a keyboard and monitor. We replace the standard hand-held keyboard and monitor worn over the eye with a Palm Pilot interface, and modify the Palm Pilot software to accept physiological input as well as traditional forms of input. The complete system can capture patterns from many kinds of user behavior, which can potentially be used to learn good predictors of user preferences. The new application presented here uses the new aaective wearable system to aid in music selection by incorporating not only stated user preferences, but also physiological variables that might be indicative of the user's present mood. AAective wearables are wearable computers equipped with physiological sensors and pattern recognition that can perceive and respond to the wearer's aaective state. These computers are designed to be comfortably worn for long periods of time, and therefore will not require constant interaction with direct input devices such as a screen or a keyboard. Instead of asking us to repeatedly click a button to indicate our preferences through menu selections, the goal of the aaective wearable is to \get to know" the wearer and their preferences through perception of the user's af-fective state PH97]. The aaect recognition algorithms and user preference prooles then act as a personal, private, perceptual user interface that can be controlled by the wearer to help manage task load or entertainment. To learn the user's natural preferences and aaective responses , it is necessary that the interface to the wearable computer be unobtrusive so that it does not cause unusual anxiety or social behavior. The natural interaction of the user and their social comfort is of primary concern for this kind of assessment. The loss of aaective communication through eye movement, facial gesture, and the social suspicion aroused by an unseen virtual environment make heads up displays (HUDs) unattractive for aaective wearables. Figure 1: Direct access to the computer has been achieved through a modiied PalmPilot interface, shown (left) as a monitor physiological sensors and (right) running the …